A PROJECTED GRADIENT AUGMENTED LAGRANGIAN APPROACH TO MULTI-OBJECTIVE TRAJECTORY PLANNING OF REDUNDANT ROBOTS

Abstract

In this paper, a multi-objective trajectory planning system is developed for redundant manipulators. This system involves kinematic redundancy resolution, as well as robot dynamics, including actuators model. The kinematic redundancy is taken into account through a secondary criterion of joint limits avoidance. The optimization procedure is performed subject to limitations on actuator torques and workspace, while passing through imposed poses. The Augmented Lagrangian with decoupling (ALD) technique is used to solve the resulting constrained non-convex and non-linear optimal control problem. Furthermore, the final state constraint is solved using a gradient projection. Simulations on a three degrees of freedom planar redundant serial manipulator show the effectiveness of the proposed system.